The way I look at ITT and PP is the difference between efficacy and effectiveness.
I have a new drug. It tastes terrible. I run a trial.
Of the subjects that take the drug as they were supposed to (PP) they do very well. Trouble is many patients don’t take the drug as required and so don’t improve. Consequently the ITT analysis fails.
If the development process is at an early stage this might be a great result. The drug has efficacy and I can fix the taste to improve compliance and hence effectiveness later.
The interaction with ITT and PP with non-inferiority trials however is more subtle.
The first simple point is a failed superiority trial cant become a successful non-inferiority trial. Pretty well full stop at that point.
But secondly PP comparisons (and not ITT) analyses are used in non-inferiority trials because essentially you are reversing the bias.
Imagine you ran a hopeless non-inferiority trial and large numbers of patients in both groups didn’t even start or comply with their treatments. The more hopeless your study was the better for showing non-inferiority (no difference) between the two treatments.
For this reason you take the PP treatment groups and compare the (hopefully) overlapping margins on them. In my hopeless study example I will fail because my small Ns of those who actually complied will produce a large CI.
Here the situation is a little bit trickier between ITT and PP because it would be expected that subject compliance isn’t the issue rather it is patient switching out of treatments. And this is informative of outcome – patients switched because they weren’t doing well. This is the bug bear of oncology trials; the ethics of the life and death make things more complicated.
SRX Price at posting:
$17.00 Sentiment: None Disclosure: Not Held
(20min delay)